Given the transition to Python 3, I don't think we should directly compare the elements in a set (though perhaps an exception should be made if they all have the same parent). Comparing their string representations seems better.
If we do sort the string representation of sets, we can reduce surprise for users by using something like natural sort (see https://github.com/SethMMorton/natsort for a library, or various code snippets available online). As for whether to do this only for doctests or for the user interface as well, I think it's nicer for the two to be the same. So I would prefer 2.2 over 2.1. David On Tue, Aug 29, 2017 at 9:25 AM, Maarten Derickx < m.derickx.stud...@gmail.com> wrote: > Option 2.3 might be to make solution 2.1 optional and dependent on the > presence of something like # unorderedset in the doctest. I actually like > this the best since it allows us to use it only in the cases where it is > really needed. So we can have the advantages of 2.1 and 2.2 depending on > the situation! > > > On Tuesday, 29 August 2017 17:44:10 UTC+2, Maarten Derickx wrote: >> >> The output of both dict and set are ordered by a display hook from >> IPython, the problem is that this sorting is done on the basis of __cmp__ >> and the default way of sorting using __cmp__ is using the id of an object >> which is not deterministic, so even though the output is sorted it is done >> on a nondeterministic basis for objects which have this nondeterministic >> __cmp__. >> >> I see basically 2 options going forward: >> >> 1. Just fix the doctests of pynormaliz themselves by making them >> deterministic. >> >> 2. Make the doctesting of sets more robust by ordering the output basted >> on the string representation instead of __cmp__ . This has two ways of >> doing it. >> 2.1 Only change the order of the output in the doctesting framework >> 2.2 Change it both in the doctesting framework and the user interface >> >> At https://trac.sagemath.org/ticket/23586 there is a partial solution >> using 2, I stopped working on this because I want feedback on wether we >> should do 2.1 or 2.2. But the progress on this ticket is slow. >> >> So I propose doing the quick fix 1 independently of 2, because this will >> remove a lot of false negatives of the patchbot on other tickets helping >> greatly in the sage development. >> >> Do people have feedback on 2.1 vs 2.2? >> Main advantage of 2.1, the user will get more meaningful output in the >> case where __cmp__ is deterministic. For example with 2.2 the user will get >> the following awkward output: >> >> sage: set(GF(11)) >> {0, 1, 10, 2, 3, 4, 5, 6, 7, 8, 9} >> >> The main advantage of 2.2 is that the doctest and the actual output of >> sage are guaranteed to agree and that the code paths for doctesting and >> actual output are more similar. >> >> My personal preference is for 2.1 since sets are inherently unordered, >> but I am open for feedback. >> >> On Friday, 18 August 2017 09:14:55 UTC+2, Daniel Krenn wrote: >>> >>> A patchbot reported two failing doctests because the result is a set and >>> the elements were in a different order, see >>> https://trac.sagemath.org/ticket/23637 >>> <https://www.google.com/url?q=https%3A%2F%2Ftrac.sagemath.org%2Fticket%2F23637&sa=D&sntz=1&usg=AFQjCNHgNMW5ruZQKIa174ovsDnStqPcDA> >>> I somehow remember that such an issue was discussed here and I thought >>> it was about the output of dict and set being sorted by a display hook. >>> Am I wrong and this is not done or even not possible? >>> >>> Best >>> >>> Daniel >>> >> -- > You received this message because you are subscribed to the Google Groups > "sage-devel" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to sage-devel+unsubscr...@googlegroups.com. > To post to this group, send email to sage-devel@googlegroups.com. > Visit this group at https://groups.google.com/group/sage-devel. > For more options, visit https://groups.google.com/d/optout. > -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at https://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.